Kalibrering av gasflödesmätare: Calibration of gas flow meters

Kalibrering av gasflödesmätare (Calibration of gas flow meters)

Jerker Delsing

Kalibrering av gasflödesmätare (Calibration of gas flow meters)

Jerker Delsing. Värme- och Kraftteknik. Lunds Tekniska Högskola

STIFTELSEN FÖR VÄRMETEKNISK FORSKNING 101 53 STOCKHOLM - TEL 08/677 25 80

Augusti 1995 ISSN 0282-3772

INSTITUTIONEN FÖR

VÄRME- OCH KRAFTTEKNIK

TEKNISKA HÖGSKOLAN I LUND

Kalibrering av gasflödesmätare

Doc. Jerker Delsing, Värme och Kraftteknik - Lunds Tekniska Högskola Sammanfattning

Vid Lunds Tekniska Högskola, institutionen for Värme och Kraftteknik, bedrivs sedan flera år forskning kring mätning av gas- och vätskeflöden. En viktigt del inom denna forskning är att kunna kalibrera mätutrustning för gasflödesmätning.

Med finansiering från Vänneforsk gjordes 1990-1991 ett försök att kalibrera en

egenutvecklad ultraljudsflödesmätare. Härvid användes en kalibrerad turbinflödesmätare som referens. I samband med dessa tester av ultraljudsgasflödesmätaren erhölls följande resultat:

• Repeterbarhet bättre än 0.5%.

• Linjariserad noggrannhet bättre än l % kan uppnås

• Nollflödesstabiliteten bättre än O.5mm/s.

De resultat som erhölls visade att den testade ultraljudsflödesmätaren hade prestanda som delvis inte matchades av den använda referensrnätaren. Att använda en referensflödesmätare för att studera hur installationseffekter påverkar gasflödesmätare har visats sig ännu svårare.

Med anledning av detta byggdes en helt ny kalibreringsutrustning. Denna baserades på en delvis ny princip som kan beskrivas som en gravimetrisk gasklocka. Detta projekt

finansierades av Nutek, Sydkraft och Vänneforsk.

Anläggningen är baserad på en omvänd gasklockeprincip. En gasmängd sugs in i gasklockan varefter vi kan väga den insugna gasmassan. Genom att härtill mäta tiden för att fylla

gasklocka kan vi erhålla ett massflöde. Genom att även mäta tryck och temperatur kanvi konvertera massflödet till volymflöde och gashastighet, detta för att enkelt kunna jämföra mot den mätare som är under test i kalibreringsriggen. Kalibreringsriggens prestanda har

uppskattats utifrån tester och delkalibreringar av ingående mätsystem. Anläggningens prestanda kan sammanfattas med:

• Kalibreringsnoggrannhet bättre än 0.17%.

• Repeterbarhet bättre än 0.1 %.

• Reynolds tal område: 0- 120.000

• Gashastighetsområde: 0-20 mls

• Kalibrering med både luft och naturgas

• Tryckområde: O- 0.5 bar

• Rördiametrar: 10 - 50 mm

• Riggen kan simulera både statiska och dynamiska installationseffekter

• Helautomatisk drift

Någon spårbarhet av den ingående mätutrustningen mot internationella standards finnsinte etablerad. Den uppskattade noggrannheten har inte verifierats mot andra laboratorier.

DEPARTMENT OF

HEAT & POWER ENGINEERING

LUND INSTITUTE OF TECHNOLOGY

Calibration of gas flow meters

by Dr. Jerker Delsing Heat and Power Engineering Lund Institute ofTechnology

Sweden Preface

At the department of Heat and Power Engineering, Lund Institute of Technology, Sweden, research is conducted within the field of gas flow measurement technology.

As a part of a development of ultrasound technology for gas flow measurement the ultrasonic flow meter was calibrated. The performance of the ultrasonic meter did not match the

performance of the calibration rig used, thus making it doubtful to judge the real performance of the ultrasonic flow meter. The calibration rig did rely on a with traceability calibrated turbine meter. The result of calibrations of the ultrasonic gas flow meter is given in part 1

"Absolute Calibration of an Ultrasonic Gas Flow Meter" of this report .

This work initiated the development of an absolute calibration rig not based on reference meter technology. A low pressure calibration rig has been designed and put into operation.

The calibration rig is based on a controlled gravimetric bell prover technology. Initial testing shows absolute accuracy of 0.17%. This work is described in part 2 "A Gas Flow Calibration Rig" of this report.

Content

Absolute Calibration of an Ultrasonic Gas Flow Meter. . . .. 1

Abstract . . . .. 1

1 Intraduction 1 2 The Sing-around method . . . .. 2

2.1 Description . . . .. 2

2.2 Limitations of metering capability . . . .. 4

2.2.1 lera flow stability . . . .. 4

2.2.2 Maxirnum flow limitations ... . . .. 5

3 Measuring equipment. . . .. 5

3.1 Description . . . .. 5

3.2 The reference meter , 6 3.3 The contra I computer . . . .. 6

3.4 Estimations of the calibration rig stability . . . .. 7

4 Measurements . . . .. 8

4.1 Sing-around versus reference . . . .. 8

4.2 lera flow measurements . . . .. 10

5 Results . . . .. 10

5.1 Absolute accuracy . . . .. 10

5.2 lera flow stability . . . .. 10

6 Conclusions . . . .. 11

7 References. . . .. 12

A Gas Flow Calibration Rig . . . .. 13

Abstract . . . .. 13

Intraduction . . . .. 13

Design criteria's . . . .. 14

Design considerations . . . .. 14

Basic rig operation and design 16 Mechanical design. . . .. 16

Operation of the rig . . . .. 19

Instrumentation. . . .. 20

Error estimations 20 First impression of use . . . .. 23

Acknowledgment . . . .. 23

References . . . .. 23

ABSOLUTE CALIBRATION ofan

ULTRASONIC GAS FLOW METER

Dr. Jerker Delsing Evert Håkansson

Abstract

The absolute calibration of an ultrasonic Sing-around gas flow meter is presented. The performance of the calibration rig used is given showing that the main problem of the rig is flow stability. The drift in flow velecity can be up

+ / - 5%

of actual flow. As reference meter a calibrated turbine meter was used. For the ultrasonic flow meter

calibration

data as repeatability, absolute

accuracy

and zero flow stability are given. The calibration data shows a repeatability of

+/- 0.5%.

After linearization we can obtain an absolute accuracy of

+ / -

1

%.

The zero flow stability

is

better than

+ / - 0.5

mm/ s.

1.

Introduction

The measurement of gas flow is of great interest from both economical and technical aspects. A new measurement technique of interest is the ultrasonic transit time technology. The use of the ultrasonic transit time technique for gas flow measurements has been shown by several authors ([1],[2],[3]).

In many industrial applications three flow meter specifications are of essential interest; accuracy, repeatability and metering range. Gas flow meters for industrial applications with an accuracy of better than 1

%

of actual flow normally shows a dynamic metering range of 1:10 or less. Unfortunately flow meters with larger dynamic ranges normally shows greater inaccuracy. A greater dynamic measuring range with maintained accuracy

is

of interest in

1

•

many applications. Earlier development of our ultrasonic gas flow meter indicates a dynamic metering range of greater than 1:40, [4].

The scope of this work has been to determine the absolute accuracy and the dynamic metering range of our own ultrasonic gas flow meter. The calibration rig used is described and estimations of its performances are made. The reference meter used was a calibrated turbine flow meter with a specified absolute accuracy of +/-1% and a repeatability of +/- 0.1% within a dynamic range of 1:20. All tests have been. made using dry air.

The calibration results for the ultrasonic gas flow meter shows that

it

is possible to achieve a repeatability of better than 0.5%. Due to the good repeatability linearization of the ultrasonic meter is possible. Thus we can c1aim a dynamic metering range of 1:12 with an accuracy better than 1%.

Unfortunately we were not able to test larger metering ranges due to limitations in the calibration rig. Further the zero flow stability has been established to better than

+/-

0.5 mm/s. This indicates that the metering range can be improved to better than 1:50 with maintained accuracy.

2. The Sing-around method

2.1. Description

For the following discussion we need a brief review of the sing-around method. We will assume a configuration as shown in figure 1.

Net~

US transducer II US transducer I

_ _ _ _

~_N_e_t

^_

-): JJJ)

Figure 1. Sing-around flow meter body.

A sing-around loop is started when we transmit a short ultrasound pulse from, say, the upstream positioned transducer. This pulse is received by the downstream transducer. The pulse is fed to the sing-around electronics, which will detect it and immediately start the transmission of the next ultrasound pulse in the same direction, thus establishing a

"sing-around"-

loop. This will go on for a number of loops. The same procedure is . subsequently repeated

in

the upstream direction. The sing-around loop will oseillate with a eertain period, t, called the sing-around period. The sing- around period depends on the speed of sound, e, of the medium between the transdueers, the transdueer distanee, L, and the fluid velocity, v. Thus, we can write the downstream and upstream sing-around periods, t1 and t2' as:

L (1) t 1=c-v'cosa

From the sing-around periods, t 1 and t 2, the fluid velocity, v 12

^is

easily found as:

(2)

To determine the fluid velocity we only need to know the transdueer distanee, L, the interrogation angle a, and measure the downstream and upstream sing-around periods, tl and t2' respectively. Unfortunately the resolution requirement on the sing-around period measurement is very high.

A sing-around period measurement resolution on the order of 1:107

is

needed to measure fluid velocities of 5 cm/s with aecuracies of 1%. For a sing-around period of 64 ms this implies an absolute time resolution of about 80 ps, (light travels 25 mm during that time). This high period measurement resolution is best obtained from a multiple period average measurement over the number of N sing-around loops.

The tested sing-around flow meter consists of three main parts: the analog part, the multiple period averaging meter and the microprocessor. The analog part is able to start, maintain and control the direction of the sing-around loop. The multiple period average meter measures the total time for N periods of the sing-around signal. Finally the microprocessor reads the period meter and calculates the fluid velocity, using an improved velocity algorithm

[5]:

3

L 1 1 1 1

v=~-·-+-·-)

2 t_n tn-l tn-2 tn-3 ⁽³⁾

The distance between the transducers is 8.8 cm and, since the interrogation angle

is

30

0 ,

the effective distance of gas and sound interaction

is

7.6 cm.

2.2. Limitations of metering capability

2.2.1. Zero flow stability

The zero flow stability is one of the most important parameters in order to establish a large dynamic range of an ultrasonic flow meter [4]. Considering a flow meter with a dynamic range from 1 to 10 m/s a zero drift error of +/- 5 mm/s will cause a maximum error of +/- 0.5%.

If

the dynamic range is increased by

10

we ring the minimum limit to 0.1 m/s the error will increase to +/- 5%. This means that the zero drift error has to be decreased by a factor 10 in order to give the same contribution to the total error (fig. 2).

Error contribution [%]

5~---,

0.8 0.9 1

Flow velocity [m/s]

0.7 0.6

0.5 0.3 0.4

0.2 0.1

2n-~:---""~----L_---_-':"'_ _

- - . J - - - l

-:A- lera flow stability 5 mm/s

3++---'\---~r---~

--o-

lera flow stabiiity 2 mm/s

~ leraflow stability 0.5 mm/s

Figure

2.

lncreasing the dynamie range ofan ultrasonic flow meter by lowering the minimum flow limit requires an improved zero flow stability in order to give thesame aror con tribu tion.

In a Sing-around flow meter the zero flow stability primarily is determined

by the signal-to-noise-ratio of the electronics. Furtherrnore temperature drift

in the receiver amplifier can induce significant zero drift errors to the flow

meter.

By using advanced electronic design including advanced filtering and triggering technique these errors are minirnized. Improvements of the s/n- ratio have also been made by optically isolating the analog and digital parts of the electronics. As a result of these design considerations a maximum zero drift error of + / - 0.5 mm/ s has been achieved.

2.2.2. Maximum flow limitations

Unfortunately the ultrasonic flow meter body design introduces volumes of gas in front of the transducers outside the main flow. These volumes eauses heavy turbulence that will impair the sound transmission through the pipe.

Due to very high velocity fluctuations the ultrasonic puls simply is "blown away". To smooth out the turbulence we use thin nets in front of the transducers. These nets simulates the pipe wall for the gas flow but are still transparent for the ultrasound. Our investigations shows that these nets allow the gas flow meter to operate at me an flow velocities of up to 14 m/s.

3. Measuring equipment

3.1. Description

The calibration rig used for these experiments is shown in figure 3. Only standard components, originally meant for use in air compressor systems or similar. are used.

Dryer Dust filter

Pressure Adjustable regulator alr-vafvs

Ultrasonic gasflow meter body

Calibrated turbine reference meter

R$232 SingAround eleetronics

250 500

lnterfaca

Computer GPIB

Frequency counter

Figure

3.

Schematic pietute of the calibraiion Tig used.

5

The airflow is generated by a central compressor system. The air is dried and then fed through a 'lOurn dust filter. After filtering the air pressure is regulated to give an airflowas eonstant as possible. To adjust the airflow a pneumatic valve controlled by a computer isused.

The ultrasonic flow meter is mounted with 25 pipe diameters straight pipe in front of and behind the meter body. The pipe diameter used is 20 mm. The reference meter is mounted with 50 pipe diameters straight pipe in front of and behind the meter body using a pipe diameter of 10 mm. This well exceeds the installation recommendations for the reference meter given by the manufacturer.

3.2. The reference meter

Since the ultrasonic gas flow meter measures the mean velocity of the gas passing through the pipe, a reference meter primarily sensibility to the same quantity is to prefer. A turbine gas flow meter, FT8-8AEE1-GEA-1 from FLOW TECHNOLOGY INC, was ehosen [5]. In this meter the flow sensitive element is a freely suspended bladed rotor positioned axially in the flow stream. The rotationaI speed of the turbine is proportional to the velocity of the fluid. The rotation of the turbine rotor generates electrical pulses in the pickup which is attached to the flow meter housing in close proximity to the turning rotor. These pulses are transfonned into a square wave signal whose frequency is proportional to the flow velocity. The actual flow velocity is then calculated by the controI computer using linear interpolation between two nearest values in the calibration data table supplied.

The flow meter was calibrated in dry air for an extended range, 1:20, and according to the manufacturer the meter accuracy is

+/-

1.0 % with a repeatability of 0.1 %. Due to restrictions set by the calibration rig the usable dynamic range of the reference meter is decreased to approximately 1:12.

3.3. The controI computer

The calibration rig is computer controlled. In addition to data acquisition the computer is used for processing and saving the data. The Sing-around flow meter communicates with the control computer via RS232 and the frequency meter, reading the reference meter value, is connected via GPIB. In order to

be

able to perform automatic measurements repeatedlya control program has been developed.

3.4. Estimations of the calibration rig stability

The valve controlling the airflow in the calibration rig shows both hysterisis and unlinearity. Together with variations in the incoming air pressure this results in problems in keeping a eonstant airflow over one measuring sequence.

To investigate the stability of the actual airflowall measured flow velocity values from the same measuring cycle can be placed i a so called Youden-plot [6]. The value measured by the reference meter gives the x-coordinate and the ultrasonic flow meter gives the y-coordinate. The graph is divided into four quadrants by two median lines.

If

only random errors are present the measuring points are expected to be equally numerous in all quadrants. The Youden-plots in figure 4 shows that a variation of the actual flow velocity of up to

+/-

5% can occur. By the use of averaging techniques however some of these problems can be eliminated.

SingAround mean velocity [m/s]

1.92 r - - - ,

SingAround mean velecity [m/s]

1 1 . 7 . , . . . - - - ,

1.87

median lines •

. _{!.~.~~ , .'} ...

• e.•

•

.. .

11.2

. .

.'

..^~ .

~ ,

1.82 +--+--+--+----'

1.82 1.87 1.92

Reference mean velocity [m/s]

10.7+--+--+---+----;

10.7 11.2 11.7

Reference mean velecity [m/s]

Figure 4. Youden-plats for mean velocities ofl.9 m/s and 11 m/s. Both plots c1early shows a variation ofthe actual flow velocity during themeasuring sequence.

7

-

4. Measurements

The absolute calibration of the Sing-around gas flow meter has been divided into

MO

experimen

ts:

1. Comparison between the flow velecity values measured by the Sing-around meter and the reference meter.

2. Zero flow stability testing of the Sing-around meter

4.1. Sing-around versus reference

To compare the accuracy of the Sing-around flow meter nine different flow velocities between

1

and

12

m/s have been investigated.

Controi voltage

M

Mean velocity [m/s1

3 1.7

3.7

I

^2.2

4.4

I

^2.4

5.1 3.5

5.8

I

^4.7

6.5 6.1

7.2 7.2

7.9 8.4

8.6 11

Table

1.

Mean flow velocity versus control voltage. The adjustable valve shows both unlinearity and hysterisis.

In

order to be able to guarantee a good reference flow value the actual flow velocity is measured la, 100 or 1000 times and then averaged. The standard deviation is calculated and then examined in order to trace instability of the airflow.

To minimize the influence of drift in the calibration rig the following

considerations are taken.

If

the standard deviation of the reference mean

velocity exceeds 1

%

the measuring sequence is considered as unstable and all

values are ignored. Under normal circumstances the standard deviation

^of

the calculated reference mean velocity is less than 0.5

%.

The mean velocity

value of the ultrasonic flow meter is calculated in the same way.

If

the standard deviation exceeds 1

%

the calculated me an velocity value is considered as false and all measured values connected to that specific measurement are ignored.

At the beginning of a measuring cycle the airflow increases asymptotic which eauses a high standard deviation. By disearding the first half of the sequence and removing glitehes eaused by electronic disturbances the standard deviation normally reaches values below 1

%.If

the standard deviation for the measurement cycle still is higher than 1

%

the reason for this is a drift in flow rate and the measurement

is

considered false and all values are ignored.

The measuring cycle for each flow value presented in table 1 is repeated more than twenty times. From these measurements the repeatability of the ultrasonic meter is estimated.

4.2. Zero flow measurements

The purpose of the zero flow measurements we re to investigate the zero flow stability of the Sing-around flow meter. The zero airflow was established by removing the connection between the calibration rig and the compressor system and plugging the pipe in both ends.

One zero flow measurement consists of 100 flow velecity values each being an average of

la,

100 or 1000 measurements. The total measuring time varies between 20 minutes and 24 hours giving a good picture of both the short- and the long time zero flow stability of the Sing-around flow meter.

5. Results

5.1. Absolute accuracy

The characteristics of the ultrasonic Sing-around flow meter has been compared to the reference meter within a flow range of 1 to 12 m/s. Flow velocities down to 0.6 m/s has also been investigated but the results of these measurements are doubtful due to high standard deviation of the reference meter value.

Between 20 and 100 measurements of the nine different flow velocities within the flow range has been made.

In

these measurements the ultrasonic flow

9

meter plus the turbine reference meter shows an repeatability of better than +/- 0.5% (fig, 5). With this repeatability

it

is possible to achieve an absolute accuracy of better than

+/ -

1

%

for the ultrasonic technology.

1

i

-

_{\ 2}

i

~ ³

•

I

4

r 5

+

^{\ 6 7 8 9}

+

^lo

..

. l! ^{~ (}

SingAround deviation[%]

(unlinearized values) 32

30 28 26 24 22 20

18 16 14 12 10 8 6 4 2 O

1 2 3 4 5 6 7 8 9 10 11

Reference mean velocity 12 [m/s]

Figure 5. The error curue for unlinearized data from the ulirasonic Sing-around gas flow meter.

Velecity data from the measuring sequences number 1 and 9 have been put in Youden-plots (discussed in 3.4) showing a significant drift of the flow velecity in the calibration rig

(d.

Hg. 4). The other measuring points showa similar behaviour.

5.2. Zero flow stability

The zero flow stability of the ultrasonic flow meter has been investigated over long and short time periods. The mean velodty was measured over a time period of approximately 10 hours showing a maximum zero drift error of less than +/- 2 mm/s. The short time zero stability has been established to better than +/- 0.5 mm/s (figure 6).

10

•

. . . .M . . . _ . . _a

--

, SingAround mean velocity [m/s]

0.01 0.008 0.006 0.004 0.002 O -0.002 -0.004 -0.006 -0.008 -0.01

O 5 10 15 20 25 [minutes]

Figure 6. The zero drift error of the ultrasonic flow meter is accomplished to less than +/-

0.5

mm/s. This is equai to +/- 0.05% error at a gas velacity of

1

m/s.

The result of the measurements done so far shows that the zero drift error of the ultrasonic Dow meter increases slightly over long time periods due to temperature drift in the receiver amplifier. To investigate this the temperature in the laboratory was logged during one measuring sequence showing a significant connection between temperature and the zero Dow stability of the ultrasonic Dow meter. These measurements indicates possibilities of maintaining the short time zero stability over very long time periods.

6. Conclusions

It

has been shown that the ultrasonic gas Dow meter developed at Lund Institute of Technology, department of Heat

&

Power Engineering has an absolute accuracy of better than +/- 1

%

over an dynamic range of 1:12. The zero flow stability is better than +/- 0.5 mm/s and the repeatability is better than +/- 0.5

%

of actual Dow.

Calibration rig and reference meter constrained the tested flow range to 1:12.

Test at Dow rates outside the Dow range of the reference meter indicates that our ultrasonic gas Dow meter is capable of achieving an absolute accuracy of

+/ -

1

%

over a Dow range larger than 1:50.

11

7. References

[1] Nolan M.E. et.al. Further Development of the British Gas Ultrasonic Flowrneter, Proc. Flow Measurement in the mid 80's, NEL Scottiand, 9-12 [une, 1986.

[2] Buess C. et.al., Design and Construction of a Pulsed Ultrasonic Air Flowrneter, IEEE Trans. on Biomedical Eng., vol. BME-33, no. 8, August 1986.

[3] Delsing J., Ultrasonic Gas Flow Meter with Corrections for large Dynamic Metering Range, Ultrasonics, vol. 27, Nov. 1989.

[4] Delsing J., The Zero Flow Performance of a Sing-around Type Flow Meter, Proc. IEEE Ultrasonic Symposium, pp 1541-1544, 1990.

[5J Delsing J., A New Velocity Algorithm for Sing-around Type Flow Meters, IEEE Trans. on UFFC, Vol UFFC-34, no. 4,

[uly

1987.

[6J EG&G Flow Technology, FT32 and smaller turbine flow meters installation, operation and maintenance manual, 1989.

[7] Youden W.J., Graphical Diagnosis of Interlaboratory Test

Results, Journal of Industrial Quality Control, pp 24-28, May 1959.

A Gas Flow Calibration Rig

BY DR. JERKER DELSING

DEPARTMENT OF HEAT AND POWER ENGINEERING LUND INSTITUTE OF TECHNOLOGY, SWEDEN ABSTRACT

A new design of a calibration rigfor gasflow is described. The rig uses gravimetric determination ofgas mass together with time measuremeni to determine mass flow. Use of precision temperature and pressure measurement enables conversion to volume flow orgas velocity. Error estimates for the calibration rigs primary mass flow reference and conversion to volume flow show an absolute error ofbetter than ^0.17%,which still has to be proven. The rig operates with air and natural gas. The system has been tested against calibrated turbine meter with god results in air. First impression of use show god results in all critical measuremenis, including mass and time determination. Thus we expect the rig to perform at least to the estimated error jigures.

The project has jointly been sponsered by Nuiek, Sydkraft AB and Värmeforsk.

Introduction

This paper will describe a new gas calibration rig design and initial testing. The rig is being constructed at the department of Heat and Power Engineering at the Lund Institute of Technology in Sweden. The rig will serve the development and investigations of gas flow meter technology where absolute contral of gas flow and calibration is essentiaI. In the future the rig can be used for calibration of small dimension low pressure consumer gas flow meters.

The main use of the rig is for testing of flow meters under controlled conditions i.e. testing of meter accuracy, dynamic range, installation effects (static and dynamic), aging, dust, etc.

Such work will aid the development of new flow meter technology as well as the general understanding of the behavior ofwell know flow meter technology. Thus the ordinary calibration rig requirements on accuracy, traceability etc. has to be fulfiled. To enable the above given tests special features concerning gas flow stability, generation ofunsteady and pulsating flow is addressed by the rig desing.

There are several calibration methods that can be used for calibration of gas flow meter. The most commonly used calibration methods are critical nozzles and bell provers. A review of such systems is found in [1]. Other more specialised designs are for example the gyroscopic weighing system at Karsto, Norway [2], and the NIST-Boulder nitrogen flow facility [3].

This paper describes a combination of a bell prover and a static gravimetric rig design.

13

•

To be able to fulfil the above mentioned special requirement offlow stability etc. a design that can be described as an active controlled gravimetric bell prover or a gravimetric piston prover was adopted. The primary reference metod is basicaly static weighing. The active piston or bell design was shosen to produce both very stable flow and controled unsteady flow.

This paper will describe the rig design and operation. Further system accuracy estimates will be given and finally some first experiences with the system will be discussed.

Design criteria 's

Since the main use of the rig is for testing of flow meters undercontrolledconditions i.e.

testing of meter accuracy, dynamic range, installation effects (static and dynamic), aging, dust, etc. the rig should be feasible for both absolute calibration as well as for generating all kind of flow situations present in real installations. Apart from specifications on accuracy, repeatability, flow stability, etc. specification for both static and dynamic installation effects was stated. The following specification was decided:

• Absolute accuracy < 0.4%

• Repeatability< 0.1%

• Reynolds number range: O - 120.000

• Gas velocity range: 0-20 mJs

• Gas velocity stability: O.OOlmJs

• Use both air and natural gas

• Pressure range: O - 0.5 bar

• Pipe diameters 10 - 50 mm

• Possibilities to introduce static installation effects. i.e. elbows valves etc.

• Possibilities to generate fluctuating flow in acontrolledway

• Fully automatic operation Design considerations

To fulfil the accuracy requirements we looked at methods used for liquids. Here static weighing often is considered to be an accurate calibration method. The use of weighing as the reference method for gas flow implies some problems. To weigh a gas volume with high resolution one either has to condensate the gas or maintain a reasonable relationship between the mass of the gas container and the mass of the gas. Since high resolution scales are

available at reasonable price our choice was to use the seeond method.

Mass resolutions of 1:10⁶are readily achievable. Assuming that it is possible to aquire 1kg gas, an accuracy of 1g is required to achive a resolution of 0.1 %. Thus the mass of the gas

container has to be less than 1000kg including all. The container has to withstand pressure of 0.5 bar and fulfil all the other requirement of the rig. To improve the mass determination accuracy an optimisation of the volume and mass of the gas container is required.

Ii_l

Positi oni ng system

Pressure resovoar

Flow meter

Figure 7. 8asic design of the calibration flow rig. The volume of the bell is activily controlled by the computer system to produce the flow. The mass, temperature and pressure of the gas is measured together with time to produce a mass or volume flow value.

The other main requirement was high gas flow stability and possibilities to introduce both static and dynamic installation effects. Using fans or compressors to produce the gas flow often introduces pulsation's and flow noise. We found that the best way to produce a stable gas flow was to establish a eonstant pressure difference against a large pressure reservoir. An ideal such is the atmosphere. When adding the desired possibilities to produce fluctuating

15

flow we did consider an active controlled bell or piston prover design. By control of the bell velocity one can maintain a very stable flow velocity or produce any desired fluctuating or oscillating flow pattern. Thus enabling the system to test dynamie behaviour of flow meters without influences on the accuracy of the reference method.

These two considerations led to abasic concept of an active controlled gravimetric bell prover design shown in figure l. Here mass flow is used as the primary reference. By adding measurements of gas pressure and temperature conversion to volume flow or gas velocity can be made. If the required measurements are made with suitable accuracy and tracability an absolute calibration facility for gas can be established.

Basic rig operation and design

The principle of operation for the calibration rig can be described with the aid of figure 2. Gas is sucked into or pressed out of a bell by in a controlled way increasing or decreasing the volume of the bell. The amount of gas sucked into or pressed out of the bell can be

determined by weighing. By accurate control of the increase/decrease of the bell volume any flow can be produced from very stable to highly pulsating flow. The most accurate way to determine the gas flow is by measuring the gas mass sucked into and subsequently pressed out from the bell tub in a two-cycle operation. Since it is not possible to continuously weigh the increase of gas mass a start stop method has to be applied. Thus the time for the suck and the press operation has to be measured. From the measurement of gas mass and time the gas massflow can be calculated. By measuring both gas temperature and pressure at the position of the flow meter under test, conversion to gas volume flow or gas velocity can be made.

Mechanical design

The piston - bell system is provided by using two concentric cylinders with slightly different diameters. The inner cylinder i sealed at bortom end using a pressure vessel end. Both

cylinders are mounted on a plate. The pocket between the two cylinders is filled with glycol.

A third cylinder sealed at top end using a pressure vessel end is placed into this pocket. Thus the glycol will form a gas seal. In this way a gas bell is constructed where the third cylinder can be moved up and down in the pocket between the inner and outer cylinder. By allowing for a gas inlet at the bortom end a lifting movement of the third cylinder will increase the volume of the bell that eauses a pressure difference between the outside pressure and the inside pressure of the bell thus giving a gas flow into the bell.

Ii

Scale Scale

1

Measure mass of empty bell

2 Suck gas into the bell

3 Measure mass of gas filled bell

Figure 8. Principle of calibration rig operation. First the empty bell is weighed.

Secondly gas is sucked into the bell. Third lifting mechanisms is disconnected and the gas filled bell is weighed again.

To be able to withstand a pressure ofO.5bar the height of the glycol seal has to be more than 4.4m. To allow for pressurised operation a total height of the glycol seal of 6.5m was

decided. The dimension of the construction is 6.5m high with a diameter of 91Omm. The bell volume can be increased by maximum 3m^3•Thus the bell can hold a gas mass of up to 3.5kg dependent on operating pressure, gas density and method of operation.

17

The cylinders of the bell were fabricated in 3mm aluminium. The main reason for choosing aluminium was to lower the mass of the bell aiming to increase the accuracy of the mass determination. The total mass of the piston - bell system comes to approximately 700 kg.

To enable a weighing of the gas entrained into the bell the whole construction has been place on a precision scale. The scale used is an Mettler Toledo weighing platform KC500-1. The scale weighing capacity of 600kg. With a preload capacity of 160kg the maximum scale load is 760. The specified reproducibility ofthe scale is better than 0.5g. Test protocols showa standard deviation ofbetter than 0.22g To achieve this reproducibility the scale has to be operated at stable temperature of ±1°C. Further the whole piston - bell construction and the scale has been built in to a housing to prevent wind from adding forces to the bell that will be sensed by the scale. Even more the scale has been placed so that vibrations from for example road traffic can be minimised.

To avoid excessive load to the scale a mechanism to unload the bell from the scale has been constructed. This mechanism lifts the bell approximately 3 mm up from the scale. Thus the bell can be operated at pressure of up to 0.5bar without darnage of the scale.

The system can be operated using air or natural gas. When using air the atmosphere is used as the gas reservoir. For natural gas the incoming pipeline pressure of 4 bar is reduced to the desired operating pressure of 50 or 200 mbar thus using the pipeline as the gas reservoir.

To perform the volume increase or decrease of the bell a lifting device was designed. The requirement here are to withstand the forces from gas pressure of maximum 0.5 bar inside the bell. Further it will be able to lift the inner bell at a very eonstant speed or apply desired fluctating velocity scheme. The movement system was build on a precision screw transmission. The helix lead of the screw is 10mm with a lead error of l Oum and an uncertainty of21lm. The screw transmission can lift and hold a load of more than 40 kN.

The transmission is driven by a motion system based on a servo controlled DC motor. The motion system has a positioning resolution ofbetter than lum for the whole length of the screw. This gives that the absolute error in the positioning and the set point of the lifting speed is only govemed by the precision of the screw transmission. The speed range of the motor allows for maximum and minimum lifting speeds ofO.33mJsand l Sum/s respectively.

These lifting speeds result in gas velocities of up to 20 m1s in 50 mm pipe and down to 0.1 m1s in 10 mm pipe.

To make accurate weighing of the bell possible a way to disconnect the lifting mechanism is necessary. This was accomplished by a cotter connection that can be controlled from a computer. Further the piping system into the bell has to be mechanical disconnected from the bell. This is done using a valve and a glycol seal close to the inlet of the bell.

Operation ofthe rig

For the prover we have two different modes of operation. The primary calibration mode will determine the gas flow from weighing and time measurement. The secondary mode will measure the gas flow by determine the gas volume in the bell from the distance the lift mechanism has move and the diameter of the bell. Combining this with a time measurement will give the gas flow.

For both modes the basic way of operating the rig can be describe with the aid offigure 2.

The starting position is with an "empty" bell and the lifting mechanism is disconnected. Then the scale is tared. Now the inlet valve is opened and the lifting of the bell is started, thus starting the gas flow. When the desired amount of gas is aquired in the bell the inlet valve is closed and position of the bell is determined from the positioning system. This gives us a measure of the volume of gas sucked into the bell that can be used as a secondary reference metod. If operated in mass mode the lifting mechanism then will compress the bell to its initial volume thus increasing the pressure in the bell. The bell will be clamped and the lifting mechanism is disconnected whereafter the mass of the gas is measured. Since the total time also is measured we can ca1culate the gas mass flow. This cycle is now reversed and the gas is pressed out from the bell then the mass again is determined for the empty bell. This two way operation gives a nice possibility to double check functionality of the rig i.e. valve leakage, scale stability etc. Further this gives a possibility to test meters that can handle reversed flow. The volume mode can be calibrated using the mass mode thus providing for a reasonable secondary standard for the rig.

The time measurement in a start stop system as this is worth considering. For the moment the time is taken from when the inlet valve opens till it is closed. Since testing of a flow meter often is made for a certain flow rate the acceleration of the flow and the deceleration of the flow should be as fast as possible. Here the piping system will serve as a large capacitor

19

preventing this. This can to some extent be compensated for by programming the controi system to use special acceleration and deceleration schemes.

To get as high absolute accuracy as possible some correction has to made. The extra measurements we apply to obtain information for corrections are:

• Temperature and pressure are measured at the location of the meter under test and at the gas intake of the bell system. Thus any differences of gas volume reminiscent in pipe between tub and flow meter before and after the test can be calculated. This enables us to correct for differences in gas volume that has gone through the meter under test and which has been measured by the bell system.

• Pressure and temperature are measured at the position of the flow meter under test.

Thus the flow meter value can be recalculated to mass flow for the comparison to the bell system.

Instrumentation

To obtain a good absolute measurement of gas flow using the above described system the following measurement equipment's are used:

l. The scale is a weighing plat form from Mettler Toledo. The calibrations performed with the bellloaded on the scale shows that resolution of 1g is achievable for mass loads of 004 - 3.5kg of gas. Here tracability to Swedish National Testing and Research Institute is established.

2. The pressure instrument is a dual channel Ruska 6222 precision pressure gauge. The pressure range of the instrument is 3.5 - 260 kPa. The absolute error of the device is better than 0.05% of full scale. Here traceability to NIST is established.

3. The temperature sensors used is precision platinum thermometers. The accuracy is specified to 0.1 DINwhich is 30mK atO'Cand 80mK at 100°C. Traceability is not yet established.

4. The time measurement is made by a timer system internai to the controi computer.

Time resolution is better than lOus.

Error estimations

The mass flow measured by the rig is ca1culate as:

m=m

_t (1)

where m is the measure has mass and t is the time. To compare this mass flow to the

measures obtained from the flow meter under test conversion to volume flow or flow velocity has to be made. Assuming 10w pressure and air or natural gas we can use the ideal gas

approximation to convert between mass m and vo1ume^V of a gas:

v ⁼⁼

^mereT

p (2)

where r is the gas eonstant T is temperature and p is the pressure. Thus the volume flow becomes:

v ⁼⁼

^mereT_pet

With knowledge of the pipe cross section area A the gas velocity becomes:

V

==

mereT peteA

(3)

(4)

By adding a term for gas leakage the overall volume flow error estimate for the rig can be written as:

I dt ^{I == I} _~ ^{I + I} d: ^{I + I} di ^{I + I} _~ ^{I + I} _~t ^{I + I} f ^I

⁽⁵⁾

Assuming a gas mass of OAkg and an scale error of 1g we can estimate a worst operating case maximal dm/m to better than 0.25%. The mass error can be improved if the rig operation is limited to cases where the acquired gas mass is as high as possible i.e. 3kg or more. Thus the mass error will decrease to 0.033%. The temperature error dT/T is estimated to better then 0.03%. The pressure error dp/p at atmospheric pressure is estimated to better 0.053%. At maximum operating pressure the pressure error decreases to 0.035%.

The estimate of the time measurement error is hard. Here considerations on valve open and close times has to be taken. This error will also be flow dependent. We have ehosen to use fully opened and closed valve as triggering points and are assurning that the trigger error is less than 5ms. With measuring times of always greater than 30s the timing error dt/t is better 0.034%. Increase of the measuring time do naturally decrease the error accordingly.

For air operation we use an air mol e mass of 28.89 g/mole. Local variation mainly due to humidity variations in the laboratory is estimated to less than 0.02%. This will cause an error in the gas eonstant term of the same amount.

Further leakage has to be estimated. The main source for leak is probably valves. Here valves approved for natural gas operation was chosen. Leaks that will impair the rig performance are backward gas leaks out from the bell through the inlet valve. Since these leaks can occur under a limited time of maximum a few minutes and a driving pressure of maximum 0.5 the

21

possible mass of gas leaked can be considered small. Thus the error due to leaks can be assumed negligible.

Finally we estimate the error eaused by the gas in piping between the meter under test and the bell. Here the difference in gas volume before and after the calibration is of interest. For this purpose we deterrnine the gas pressure and temperature both at the position of the meter under test and close to the bell. The total gas volume in the pipings is at maximum 0.02 nr', The change of this volume is deterrnined from measured values of temperature and pressure.

With the accuracy stated above the rig error due to this come to less than 0.005% i.e.

negligible.

In total an estimated worst case error comes to 0.17%. The largest single source of error is here the pressure deterrniantion, closly followed by mass temperature and time. Of these the time and temperature measurement probably are easiest to improve.

4

3.5^I-

-

~ 3 ^I-

• ^-

vo

C_<l)

E 2.5^I-

-

<l)

...::l vo'"

<l)

E 2

- -

'G~ Qo

> 1.5

- -

' -o

"O

'"

•

<l)

... _I

- .. ^-

lZlo-

0.5

-

• •

• • • • •

o

5 10 15 20 25 30 35 40 45 50

Velocitylm/s]

Figure 9. The spread of five velocity measurements performed for 12 different gas velocities using a turbine meter against the calibration rig. Most of the spread seen can be attributed to start stop errors of the turbine meter.

First impression of use

The rig has recently been taken into operation. For the moment operation is done with air.

The flow meter section of the rig does provide more than 300D of pipe in front of the meter and plenty of space for the introduction of flow disturbances like elbows, etc.

Infigure 9 a test run using an calibrated turbine meter is shown. The comparison shows the spread of the turbine meter for five measurements at different flow velocities, The main reason for the spread is start stop errors. i.e, the turbine will not start below a certain velecity and stops below another velocity that is larger than zero. This phenomena is exaggerated at low flow velocities as can be seen in figure 9.

The rig has not been certified against any standardisation organisation. Further no round robin tests has been performed yet. Since such test frequently are run among the european laboratories we expect to participate in such a test in the future.

ACKNOWLEDGMENT

I like to express my sincere gratttude to Nils Widing for his support durmg this project. Further the work by Ulf Nilsson in finishing the construction of the rig is much appreciated. The

project was jointly sponsord by Sydkraft AB, Värmeforsk and the Swedish National Board for Industrial and Technical Development, NUTEK.

References

[1] Mattingly G.E., Primary calibrators, reference and transfer standards, in Developments in Flow Measurement-l, R.W.W. Scott (ed.), Appl. Sci. Publishers Ltd., Chapter 2, p.31-71.

(ISBN 0-85334-976-2),1982

[2] Velde B., A gyroscopic Weighing system for primary calibration of sonic nozzles, Int. conf.

Mass Flow Measurement Direct^& Indirect,mc,London, Feb 1989.

[3] McFaddin S., BrermanJ.A.,Sindt C.F., The pression and accuracy of mass flow measurement in the NIST-Boulder nitrogen flow facility, Int. conf. Mass Flow Measurement Direct&Indirect,

mc,London, Feb 1989.

23

VÄRMEFORSK Service AB 101 53 Stockholm

Värmeforsk är ett organ för industrisamverkan inom värmeteknisk forskning och utveckling. Forskningsprogrammet är tillämpningsinriktat och fokuseras på energi- och processindustriernas behov och problem. Bakom Värmeforsk står följande huvudmän:

N Elforsk

N Svenska Fjärrvärmeföreningen N Skogsindustrierna

N Övrig Industri

Värmeforsk samarbetar med Statens Energimyndighet.